Search visibility drops
Compare priority-page indexing, useful impressions and the pages that start productive journeys.
Next decision: Check whether the change affects all pages or one template before editing titles or producing new content.
34 indicators for a useful digital service.
A shared framework connecting visibility, experience, conversion, editorial quality, efficiency and resilience. Each indicator explains what it informs, how to calculate it and how to read it without treating an average as a universal truth.
Method
A decision framework, not a ranking.
These reference points are not market averages. Use them to establish a baseline, follow change and ask better questions. LCP, INP and CLS thresholds follow the official Core Web Vitals definitions; adapt other targets to the service, audience and level of risk.
Begin with three to five indicators directly connected to a decision. Add diagnostic measures only when they help explain a change.
34documented indicators
7complementary dimensions
CSVopen downloadable data
Framework
Choose an indicator for the decision.
Tables remain usable with a keyboard and on small screens. Scroll horizontally when needed.
Visibility and acquisition
| Indicator | Decision informed | Calculation | How to read it | Reference |
|---|---|---|---|---|
| Useful organic impressions | Check whether priority pages appear for searches connected to the offer. | Impressions for priority queries and pages during the period. Collection: Search Console export: retain query, page, device and period. | Compare by intent and landing page, not only at site level. | Contextualised monthly trend |
| Organic click-through rate | Assess the fit between a query, title and search excerpt. | Organic clicks ÷ organic impressions × 100. Collection: Use the same export for clicks and impressions with identical filters. | Compare similar queries and positions rather than isolated averages. | Change by query group |
| Contributing landing pages | Identify content that begins genuinely useful journeys. | Landing pages that generated at least one useful action. Collection: Verified conversion events and landing pages with explicit attribution rules. | A low-traffic page can still be valuable when enquiries are qualified. | Growth of the useful portfolio |
| Intent coverage | Find important questions without a satisfactory answer. | Priority intents covered by a useful page ÷ all priority intents. Collection: Editorial inventory of priority questions and pages answering them. | Measure coverage quality before increasing editorial volume. | Documented internal target |
| Indexing of priority pages | Check that pages intended for search can appear in the index. | Confirmed indexed priority pages ÷ published priority pages × 100. Collection: Indexing reports and URL inspection; a sitemap alone is insufficient. | Separate deliberate exclusions from issues affecting a page template. | Internal target for indexable pages |
| Orphan priority pages | Find important content without an internal navigation route. | Priority pages with no incoming HTML link from another site page. Collection: HTML-link crawl compared with the complete public-page inventory. | Compare the sitemap, crawl and rendered links before fixing. | Target: no orphan priority page |
Experience and conversion
| Indicator | Decision informed | Calculation | How to read it | Reference |
|---|---|---|---|---|
| Task success rate | Learn whether a person completes the action they came to perform. | Successful observed test sessions ÷ all observed sessions × 100. Collection: Observed tests: count successful and total attempts for the same task. | Define one precise task, such as understanding an offer or making contact. | Compare before and after improvement |
| Contact completion rate | Detect friction in a form or appointment journey. | Successful submissions ÷ form starts × 100. Collection: Start/success events; check duplicate submissions and applicable consent. | Segment by device and inspect validation errors. | Continuous improvement |
| Qualified enquiries | Connect the website to commercial or operational value. | Enquiries matching defined criteria ÷ enquiries received × 100. Collection: Enquiry qualification records using stable criteria and restricted access. | Agree the criteria with the teams that handle enquiries. | Target specific to the activity |
| Time to key information | Measure the effort required to reach an important answer. | Median observed time between arrival and the information sought. Collection: Timing of comparable task tests; record failures too. | Complement analytics with short user tests. | Reduction without loss of understanding |
| Blocking form errors | Find fields or validation rules that prevent a useful enquiry. | Attempts blocked by validation ÷ observed attempts × 100. Collection: De-identified error logs and validation tests; exclude sensitive field contents. | Separate input mistakes, technical failures and exits; test on mobile. | Reduction after removing obstacles |
| Offer comprehension | Check that a visitor can explain what is offered and for whom. | Participants who paraphrase the offer correctly ÷ participants tested × 100. Collection: Comprehension questions after reading, using a prepared response rubric. | Use an open question after a short reading period without prompting. | Compare before and after rewriting |
Performance and technical quality
| Indicator | Decision informed | Calculation | How to read it | Reference |
|---|---|---|---|---|
| Largest Contentful Paint (LCP) | Measure how quickly the main visible content loads. | 75th percentile of real visits, split by mobile and desktop. Collection: CrUX or RUM field data: state 75th percentile, device and period. | Review the server, LCP resource, critical CSS and rendering. | Good: ≤ 2.5 s |
| Interaction to Next Paint (INP) | Measure overall responsiveness to interactions. | 75th percentile of real visits, split by mobile and desktop. Collection: Field interaction data; a test without interaction does not measure actual INP. | Inspect long tasks, third-party scripts and DOM updates. | Good: ≤ 200 ms |
| Cumulative Layout Shift (CLS) | Measure unexpected visual movement. | 75th percentile of real visits, split by mobile and desktop. Collection: Field stability data; confirm causes using a laboratory trace. | Reserve space for media, components and injected content. | Good: ≤ 0.1 |
| Crawlable internal links | Support discovery of important pages and understandable journeys. | Valid HTML internal links ÷ all checked internal links × 100. Collection: Crawl rendered HTML: actual destination and link accessibility. | Fix orphan pages, errors and ambiguous anchors first. | Target: 100% for priority pages |
| Expected server responses | Find errors and unexpected redirects on essential journeys. | Priority requests returning the expected response ÷ priority requests tested × 100. Collection: Check requests against the expected status for each URL, including redirects and missing pages. | Separate server errors, missing pages and intended redirects. | Target: 100% on tested journeys |
| Primary media weight | Limit transfer and loading delay caused by the dominant image or video. | Bytes transferred for the primary media on each priority template. Collection: Network panel for the media actually received for the chosen screen and format. | Check format, dimensions, mobile variant and loading priority. | Internal budget by template and device |
Editorial quality and trust
| Indicator | Decision informed | Calculation | How to read it | Reference |
|---|---|---|---|---|
| Evidence coverage | Check whether important claims can be verified. | Supported priority claims ÷ checked priority claims × 100. Collection: Review verifiable claims and the evidence actually available. | Evidence can be a primary source, method, attributed example or explicit limitation. | Internal target by content type |
| Freshness of critical sources | Find content whose references or rules may have changed. | Critical sources reviewed within the set interval ÷ all critical sources × 100. Collection: Source register separating event date, publication date and review date. | Review frequency depends on risk: regulation, prices and tools change quickly. | 100% within the defined review cycle |
| Content-assisted journeys | Measure content contribution without requiring an immediate conversion. | Useful journeys that used help content ÷ all useful journeys observed × 100. Collection: Instrumented journeys or user feedback; state attribution limitations. | Study sequences and feedback without giving automatic credit to the last page. | Trend by priority journey |
| Complete editorial attribution | Let readers identify who is responsible for a piece of content and its context. | Editorial pages with publisher, date and applicable sources ÷ pages checked × 100. Collection: Page inventory with publisher, meaningful date and sources where needed. | Require a source where a checkable claim depends on one. | Internal target by page type |
| Consistency across language versions | Prevent translated pages from giving conflicting facts or routes. | Reviewed page pairs without critical divergence ÷ pairs checked × 100. Collection: Review FR/EN pairs: facts, links, dates, next action and language mapping. | Check figures, links, dates, calls to action and hreflang equivalents. | Target: no critical divergence |
Efficiency and resilience
| Indicator | Decision informed | Calculation | How to read it | Reference |
|---|---|---|---|---|
| Transferred weight per key page | Limit the data and energy needed for common journeys. | Bytes transferred in a representative load, split by resource type. Collection: Network captures with cold then warm cache under identical conditions. | Compare similar templates and inspect images, fonts, scripts and unnecessary responses. | Internal budget per template |
| Third-party service budget | Control the impact, dependency and risk of external scripts. | Count, weight, requests and CPU time of authorised third-party services. Collection: Third-party inventory with requests, bytes, CPU and owner. | Review each service for usefulness, consent, performance and ownership. | Approved list reviewed quarterly |
| Successful restore rate | Verify that the service and its data can actually be restored. | Restore tests completed within the target time ÷ planned tests × 100. Collection: Planned restoration records: successes, failures and exercises not carried out. | Record duration, lost data, dependencies and gaps found during every exercise. | Target: 100% of planned tests |
| Static cache reuse | Reduce repeated transfers of unchanged media, styles and scripts. | Static requests served from cache ÷ eligible static requests × 100. Collection: Network panel: distinguish browser cache, CDN and server responses. | Segment by resource and verify updates after publication. | Improvement without stale content |
| Time to restore service | Measure the ability to restore a journey after an incident. | Time from detection to verified restoration, per incident or exercise. Collection: Incident log: detection, verified recovery and definition of incident end. | Compare with service criticality and document slow stages. | Internal target by criticality |
Accessibility and inclusion
| Indicator | Decision informed | Calculation | How to read it | Reference |
|---|---|---|---|---|
| Keyboard journey success | Check that essential actions remain possible without a mouse. | Essential journeys completed with a keyboard ÷ essential journeys tested × 100. Collection: Complete keyboard tests of essential tasks, including focus and absence of traps. | Review focus order, visibility, menus, forms and dialogs. | Target: 100% of essential journeys |
| Informative images described | Make essential visual information available in text. | Informative images with suitable alternatives ÷ informative images checked × 100. Collection: Human review of the information conveyed by each image; do not count decorative images. | Judge relevance in context, not just the presence of an alt attribute. | Target: 100% of informative images |
Data quality and comparisons
| Indicator | Decision informed | Calculation | How to read it | Reference |
|---|---|---|---|---|
| Datasets with documented reuse context | Check that used data retains provenance and rights. | Used datasets with documented publisher, licence, period and version ÷ used datasets × 100. Collection: Data register: publisher, dataset and resource URLs, licence, period, version and transformations. | A portal URL alone is insufficient; inspect the actual resource used. | Internal target: every used dataset is documented |
| Unmatched records | Detect coverage loss during a data join. | Starting-source records without a valid match ÷ starting-source records × 100. Collection: Join log recording starting-source size and unique, multiple and absent matches. | Retain unmatched records; examine multiple matches and validation rules separately. | Internal threshold suited to the decision |
| Unconfirmed cost items | Expose unknowns before comparing proposals. | Number of required cost items with unconfirmed amounts or assumptions, per solution. Collection: Proposal comparison: setup, recurring charges, internal effort, change and exit, with each amount’s source and status. | Do not treat an unknown cost as zero. Separate unknown, confirmed zero and estimate. | Target: unknowns explicit before deciding |
| Retested journey corrections | Confirm that a correction resolves the observed obstacle. | Priority corrections retested with a documented result ÷ priority corrections declared complete × 100. Collection: Test register: task, initial obstacle, change, date and repeat-test result. | Use the same scenario and comparable conditions; record new obstacles. | Internal target: every priority correction is retested |
How to use it
Build a dashboard in four steps.
The quality of the discussion matters more than the sophistication of the tool.
- 01 · Outcome
- State the expected result and connected decision.
- 02 · Definition
- Set the calculation, source, segments and limitations.
- 03 · Context
- Choose a relevant comparison period and annotate changes.
- 04 · Action
- Assign the next step, owner and validation.
From signal to action
Six situations, and the measures worth combining.
A single metric rarely explains a problem. Use these examples to formulate a question, gather contrasting evidence and assign the next test.
People abandon a form
Compare form completion, blocking validation errors and task success by device.
Next decision: Watch a few real attempts. Fix unclear questions and error recovery before adding another acquisition campaign.
A page feels slow
Read field LCP, INP and CLS separately for mobile and desktop, then inspect the main media weight.
Next decision: Find the affected template and resource before compressing every asset or changing unrelated code.
A guide may be outdated
Review critical-source freshness, evidence coverage and the content owner.
Next decision: Correct the affected claim and its language equivalent; change the review date only after checking the material.
A release causes incidents
Check expected server responses, restoration tests and time to recover.
Next decision: Document the failure and recovery path, then improve the release checklist or backup procedure.
A task excludes people
Combine keyboard task success with a review of informative image alternatives and form feedback.
Next decision: Test the complete journey with people and assistive technology where possible; a count of missing alt attributes is insufficient.
Go further
From measurement to improvement.
Select useful KPIs
Build a measurement plan linked to objectives and responsibilities.
Read the KPI guide →Improve Core Web Vitals
Diagnose LCP, INP and CLS with field and laboratory data.
Read the performance guide →Prioritise technical SEO
Connect technical findings to pages and journeys that matter.
Read the audit guide →Sources and limits
A transparent, revisable framework.
Core Web Vitals thresholds are documented by web.dev. Crawling and content guidance follows Google Search Central. Efficiency and resilience targets must be contextualised; they are not universal guarantees.
Framework version: 6 October 2026. Suggest a correction through the contact page.
Resources for this project
Affiliate marketing guide
World Affiliate Guide
For an affiliate-funded website, review commission, attribution and conversion concepts. Compare the calculation conditions behind each metric rather than isolated rates.
Tourism technology guide
Performance Voyage
For a travel booking journey, identify dependencies between business tools, distribution and payment. Document where journey measurements begin and end.
Website in French
Sites from the AR ecosystem, selected for their relevance to this topic. Check scope, terms and current offers with each publisher. Our selection method.